Add LRU cache support.

Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

README.md

@@ -155,6 +155,7 @@ You can still use the non-prefixed environment variables in the spirit of the ba
 | `CACHE_TTL`               | The cache TTL                           | `60`                       |
 | `CACHE_MAX_MEMORY_SIZE`   | Maximum memory size for cache in MB     | `100`                      |
 | `CACHE_MAX_ENTRIES`       | Maximum number of entries in cache      | `10000`                    |
+| `CACHE_USE_LRU`           | Use LRU eviction algorithm (see [Cache Eviction](#cache-eviction-algorithms)) | `false`    |
 | `CACHE_PER_USER_DISABLED` | **⚠️ SECURITY**: Disable per-user cache isolation | `false` (**DO NOT** set to `true` in multi-user apps) |
 | `ENABLE_REDIS_CACHE`      | Enable distributed Redis cache          | `false`                    |
 | `CACHE_REDIS_URL`         | URL to redis server / cluster endpoint  | `localhost:6379`           |
@@ -439,6 +440,32 @@ These features ensure the cache runs efficiently even under high load and with l
 Since version `0.5.30` the cache is gzipped in the memory, which should optimise the memory usage quite significantly.
 Since version `0.15.48` the you can also use the distributed Redis cache.
 
+#### Cache Eviction Algorithms
+
+The proxy supports two cache eviction strategies:
+
+**Standard (default):** Uses Go's `sync.Map` with approximate eviction. When memory limits are reached, entries are evicted based on iteration order (pseudo-random). This is memory-efficient and has excellent concurrent read performance.
+
+**LRU (Least Recently Used):** Uses a proper LRU algorithm with a linked list to track access order. When limits are reached, the least recently accessed entries are evicted first. Enable with `CACHE_USE_LRU=true`.
+
+| Feature | Standard | LRU |
+|---------|----------|-----|
+| Eviction order | Pseudo-random | Least recently used |
+| Read performance | Excellent | Good |
+| Memory tracking | Approximate | Precise |
+| Best for | High read throughput | Cache hit optimization |
+
+*LRU cache configuration:*
+```bash
+GMP_ENABLE_GLOBAL_CACHE=true
+GMP_CACHE_TTL=300
+GMP_CACHE_USE_LRU=true
+GMP_CACHE_MAX_MEMORY_SIZE=200
+GMP_CACHE_MAX_ENTRIES=5000
+```
+
+Use LRU when cache hit rate is critical and you want to ensure frequently accessed data stays cached. Use Standard (default) for maximum read throughput with less memory overhead.
+
 #### Read-only endpoint
 
 You can now specify the read-only GraphQL endpoint by setting the `HOST_GRAPHQL_READONLY` environment variable. The default value is empty, preventing the proxy from using the read-only endpoint for the queries and directing all the requests to the main endpoint specified as `HOST_GRAPHQL`. If the `HOST_GRAPHQL_READONLY` is set, the proxy will use the read-only endpoint for the queries with the `query` type and the main endpoint for the `mutation` type queries. Format of the read-only endpoint is the same as `HOST_GRAPHQL` endpoint, for example `http://localhost:8080/`.

admin/dashboard.html

@@ -1277,15 +1277,34 @@
             stateEl.classList.remove('loading');
 
             let badgeClass = 'badge-info';
-            if (data.state === 'closed') badgeClass = 'badge-success';
-            else if (data.state === 'open') badgeClass = 'badge-danger';
-            else if (data.state === 'half-open') badgeClass = 'badge-warning';
+            let stateText = data.state || 'unknown';
 
-            stateEl.innerHTML = `<span class="badge ${badgeClass}">${data.state || 'Unknown'}</span>`;
+            // For cluster mode, determine state from instance counts
+            if (data.instances_open !== undefined) {
+                // Cluster mode data
+                if (data.instances_open > 0) {
+                    stateText = `${data.instances_open} open`;
+                    badgeClass = 'badge-danger';
+                } else if (data.instances_halfopen > 0) {
+                    stateText = `${data.instances_halfopen} half-open`;
+                    badgeClass = 'badge-warning';
+                } else if (data.instances_closed > 0) {
+                    stateText = `${data.instances_closed} closed`;
+                    badgeClass = 'badge-success';
+                }
+            } else {
+                // Single instance mode
+                if (data.state === 'closed') badgeClass = 'badge-success';
+                else if (data.state === 'open') badgeClass = 'badge-danger';
+                else if (data.state === 'half-open') badgeClass = 'badge-warning';
+            }
+
+            stateEl.innerHTML = `<span class="badge ${badgeClass}">${stateText}</span>`;
 
             document.getElementById('cb-enabled').textContent = data.enabled ? 'Yes' : 'No';
 
             if (data.counts) {
+                // Single instance mode with detailed counts
                 document.getElementById('cb-total-requests').textContent =
                     (data.counts.requests || 0).toLocaleString();
                 document.getElementById('cb-total-successes').textContent =
@@ -1296,6 +1315,14 @@
                     (data.counts.consecutive_successes || 0).toLocaleString();
                 document.getElementById('cb-consecutive-failures').textContent =
                     (data.counts.consecutive_failures || 0).toLocaleString();
+            } else if (data.instances_open !== undefined) {
+                // Cluster mode - show instance distribution instead
+                const total = (data.instances_open || 0) + (data.instances_closed || 0) + (data.instances_halfopen || 0);
+                document.getElementById('cb-total-requests').textContent = total + ' instances';
+                document.getElementById('cb-total-successes').textContent = (data.instances_closed || 0).toLocaleString();
+                document.getElementById('cb-total-failures').textContent = (data.instances_open || 0).toLocaleString();
+                document.getElementById('cb-consecutive-successes').textContent = '--';
+                document.getElementById('cb-consecutive-failures').textContent = '--';
             }
 
             if (data.config) {
@@ -1307,23 +1334,31 @@
         function updateCoalescing(data) {
             document.getElementById('coalescing-rate').textContent =
                 (data.backend_savings_pct || 0).toFixed(1) + '%';
-            document.getElementById('coalescing-total').textContent =
-                (data.total_requests || 0).toLocaleString();
-            document.getElementById('coalescing-primary').textContent =
-                (data.primary_requests || 0).toLocaleString();
-            document.getElementById('coalescing-coalesced').textContent =
-                (data.coalesced_requests || 0).toLocaleString();
+
+            // Handle both single instance (total_requests) and cluster mode (total_coalesced + total_primary)
+            const totalRequests = data.total_requests ||
+                ((data.total_coalesced_requests || 0) + (data.total_primary_requests || 0));
+            document.getElementById('coalescing-total').textContent = totalRequests.toLocaleString();
+
+            // Handle both single instance and cluster mode field names
+            const primaryRequests = data.primary_requests || data.total_primary_requests || 0;
+            document.getElementById('coalescing-primary').textContent = primaryRequests.toLocaleString();
+
+            const coalescedRequests = data.coalesced_requests || data.total_coalesced_requests || 0;
+            document.getElementById('coalescing-coalesced').textContent = coalescedRequests.toLocaleString();
+
             document.getElementById('coalescing-savings').textContent =
                 (data.backend_savings_pct || 0).toFixed(1) + '%';
         }
 
         function updateRetryBudget(data) {
-            document.getElementById('retry-tokens').textContent =
-                data.current_tokens || '--';
-            document.getElementById('retry-current-tokens').textContent =
-                data.current_tokens || '--';
-            document.getElementById('retry-max-tokens').textContent =
-                data.max_tokens || '--';
+            // Use explicit undefined check to handle 0 values correctly
+            const currentTokens = data.current_tokens !== undefined ? data.current_tokens : '--';
+            const maxTokens = data.max_tokens !== undefined ? data.max_tokens : '--';
+
+            document.getElementById('retry-tokens').textContent = currentTokens;
+            document.getElementById('retry-current-tokens').textContent = currentTokens;
+            document.getElementById('retry-max-tokens').textContent = maxTokens;
             document.getElementById('retry-total').textContent =
                 (data.total_attempts || 0).toLocaleString();
             document.getElementById('retry-denied').textContent =
@@ -1333,13 +1368,17 @@
         }
 
         function updateWebSocket(data) {
-            document.getElementById('ws-connections').textContent =
-                data.active_connections || 0;
+            // Handle both single instance (active_connections) and cluster mode (total_connections)
+            const connections = data.active_connections !== undefined ? data.active_connections :
+                               (data.total_connections !== undefined ? data.total_connections : 0);
+            document.getElementById('ws-connections').textContent = connections;
         }
 
         function updateConnections(data) {
-            document.getElementById('pool-connections').textContent =
-                data.active_connections || 0;
+            // Handle both single instance (active_connections) and cluster mode (total_active)
+            const connections = data.active_connections !== undefined ? data.active_connections :
+                               (data.total_active !== undefined ? data.total_active : 0);
+            document.getElementById('pool-connections').textContent = connections;
         }
 
         async function resetCoalescing() {

admin_dashboard.go

@@ -429,7 +429,7 @@ func (ad *AdminDashboard) getWebSocketStats(c *fiber.Ctx) error {
 
 // clearCache clears the cache
 func (ad *AdminDashboard) clearCache(c *fiber.Ctx) error {
-	// TODO: Implement cache clearing
+	libpack_cache.CacheClear()
 	return c.JSON(map[string]interface{}{
 		"success": true,
 		"message": "Cache cleared successfully",

admin_dashboard_test.go

@@ -214,6 +214,7 @@ func TestAdminDashboard_GetCacheStats(t *testing.T) {
 			CacheRedisEnable      bool
 			CacheMaxMemorySize    int
 			CacheMaxEntries       int
+			CacheUseLRU           bool
 			GraphQLQueryCacheSize int
 			PerUserCacheDisabled  bool
 		}{
@@ -221,6 +222,7 @@ func TestAdminDashboard_GetCacheStats(t *testing.T) {
 			CacheTTL:             60,
 			CacheMaxMemorySize:   100,
 			CacheMaxEntries:      10000,
+			CacheUseLRU:          false,
 			PerUserCacheDisabled: false,
 		},
 	}

cache/cache.go

@@ -27,6 +27,7 @@ type CacheConfig struct {
 	Memory struct {
 		MaxMemorySize int64 `json:"max_memory_size"` // Maximum memory size in bytes
 		MaxEntries    int64 `json:"max_entries"`     // Maximum number of entries
+		UseLRU        bool  `json:"use_lru"`         // Use LRU eviction algorithm instead of random eviction
 	}
 	TTL                  int  `json:"ttl"`
 	IncludeUserContext   bool `json:"include_user_context"`    // Include user ID and role in cache key
@@ -96,16 +97,6 @@ func CalculateHash(c *fiber.Ctx, userID string, userRole string) string {
 	return strutil.Md5(cacheKeyData)
 }
 
-// CalculateHashLegacy generates a cache hash using only the request body (DEPRECATED).
-// This function exists for backward compatibility only and should NOT be used
-// in production multi-user applications as it creates a security vulnerability
-// where users can see each other's cached data.
-//
-// Deprecated: Use CalculateHash with user context instead.
-func CalculateHashLegacy(c *fiber.Ctx) string {
-	return strutil.Md5(c.Body())
-}
-
 func EnableCache(cfg *CacheConfig) {
 	if cfg.Logger == nil {
 		cfg.Logger = libpack_logger.New()
@@ -134,34 +125,41 @@ func EnableCache(cfg *CacheConfig) {
 			cfg.Client = libpack_cache_redis.NewCacheWrapper(redisClient, cfg.Logger)
 		}
 	} else {
+		// Calculate memory and entry limits
+		maxMemory := cfg.Memory.MaxMemorySize
+		if maxMemory <= 0 {
+			maxMemory = libpack_cache_memory.DefaultMaxMemorySize
+		}
+
+		maxEntries := cfg.Memory.MaxEntries
+		if maxEntries <= 0 {
+			maxEntries = libpack_cache_memory.DefaultMaxCacheSize
+		}
+
+		cacheType := "standard"
+		if cfg.Memory.UseLRU {
+			cacheType = "LRU"
+		}
+
 		cfg.Logger.Debug(&libpack_logger.LogMessage{
 			Message: "Using in-memory cache",
 			Pairs: map[string]interface{}{
-				"max_memory_size_bytes": cfg.Memory.MaxMemorySize,
-				"max_entries":           cfg.Memory.MaxEntries,
+				"type":                  cacheType,
+				"max_memory_size_bytes": maxMemory,
+				"max_entries":           maxEntries,
 			},
 		})
 
-		// Use memory size and entry limits if configured, otherwise use defaults
-		if cfg.Memory.MaxMemorySize > 0 || cfg.Memory.MaxEntries > 0 {
-			maxMemory := cfg.Memory.MaxMemorySize
-			if maxMemory <= 0 {
-				maxMemory = libpack_cache_memory.DefaultMaxMemorySize
-			}
-
-			maxEntries := cfg.Memory.MaxEntries
-			if maxEntries <= 0 {
-				maxEntries = libpack_cache_memory.DefaultMaxCacheSize
-			}
-
+		if cfg.Memory.UseLRU {
+			// Use LRU cache with proper eviction algorithm
+			cfg.Client = libpack_cache_memory.NewLRUMemoryCache(maxMemory, maxEntries)
+		} else {
+			// Use standard sync.Map-based cache
 			cfg.Client = libpack_cache_memory.NewWithSize(
 				time.Duration(cfg.TTL)*time.Second,
 				maxMemory,
 				maxEntries,
 			)
-		} else {
-			// Backward compatibility
-			cfg.Client = libpack_cache_memory.New(time.Duration(cfg.TTL) * time.Second)
 		}
 	}
 	config = cfg
@@ -271,6 +269,9 @@ func CacheGetQueries() int64 {
 }
 
 func CacheClear() {
+	if !IsCacheInitialized() {
+		return
+	}
 	config.Client.Clear()
 	cacheStats = &CacheStats{}
 }

cache/memory/lru_memory_cache.go

@@ -279,3 +279,8 @@ func (c *LRUMemoryCache) GetMemoryUsage() int64 {
 func (c *LRUMemoryCache) GetMaxMemorySize() int64 {
 	return c.maxMemorySize
 }
+
+// CountQueries returns the number of entries in the cache
+func (c *LRUMemoryCache) CountQueries() int64 {
+	return atomic.LoadInt64(&c.currentCount)
+}

cache/memory/lru_memory_cache_test.go

@@ -0,0 +1,343 @@
+package libpack_cache_memory
+
+import (
+	"fmt"
+	"sync"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/suite"
+)
+
+type LRUMemoryCacheTestSuite struct {
+	suite.Suite
+}
+
+func TestLRUMemoryCacheTestSuite(t *testing.T) {
+	suite.Run(t, new(LRUMemoryCacheTestSuite))
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestNewLRUMemoryCache() {
+	cache := NewLRUMemoryCache(1024*1024, 100) // 1MB, 100 entries
+	suite.NotNil(cache)
+	suite.Equal(int64(0), cache.CountQueries())
+	suite.Equal(int64(0), cache.GetMemoryUsage())
+	suite.Equal(int64(1024*1024), cache.GetMaxMemorySize())
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestSetAndGet() {
+	cache := NewLRUMemoryCache(1024*1024, 100)
+
+	// Set a value
+	cache.Set("key1", []byte("value1"), 5*time.Second)
+
+	// Get the value
+	val, found := cache.Get("key1")
+	suite.True(found)
+	suite.Equal([]byte("value1"), val)
+
+	// Get non-existent key
+	val, found = cache.Get("nonexistent")
+	suite.False(found)
+	suite.Nil(val)
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestUpdateExisting() {
+	cache := NewLRUMemoryCache(1024*1024, 100)
+
+	cache.Set("key1", []byte("value1"), 5*time.Second)
+	cache.Set("key1", []byte("value2"), 5*time.Second)
+
+	val, found := cache.Get("key1")
+	suite.True(found)
+	suite.Equal([]byte("value2"), val)
+	suite.Equal(int64(1), cache.CountQueries())
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestDelete() {
+	cache := NewLRUMemoryCache(1024*1024, 100)
+
+	cache.Set("key1", []byte("value1"), 5*time.Second)
+	suite.Equal(int64(1), cache.CountQueries())
+
+	cache.Delete("key1")
+	suite.Equal(int64(0), cache.CountQueries())
+
+	val, found := cache.Get("key1")
+	suite.False(found)
+	suite.Nil(val)
+
+	// Delete non-existent key should not panic
+	cache.Delete("nonexistent")
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestClear() {
+	cache := NewLRUMemoryCache(1024*1024, 100)
+
+	cache.Set("key1", []byte("value1"), 5*time.Second)
+	cache.Set("key2", []byte("value2"), 5*time.Second)
+	cache.Set("key3", []byte("value3"), 5*time.Second)
+	suite.Equal(int64(3), cache.CountQueries())
+
+	cache.Clear()
+	suite.Equal(int64(0), cache.CountQueries())
+	suite.Equal(int64(0), cache.GetMemoryUsage())
+
+	_, found := cache.Get("key1")
+	suite.False(found)
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestExpiration() {
+	cache := NewLRUMemoryCache(1024*1024, 100)
+
+	cache.Set("key1", []byte("value1"), 100*time.Millisecond)
+
+	// Should exist immediately
+	val, found := cache.Get("key1")
+	suite.True(found)
+	suite.Equal([]byte("value1"), val)
+
+	// Wait for expiration
+	time.Sleep(150 * time.Millisecond)
+
+	// Should be expired
+	val, found = cache.Get("key1")
+	suite.False(found)
+	suite.Nil(val)
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestEvictionByCount() {
+	cache := NewLRUMemoryCache(1024*1024, 3) // Max 3 entries
+
+	cache.Set("key1", []byte("value1"), 5*time.Second)
+	cache.Set("key2", []byte("value2"), 5*time.Second)
+	cache.Set("key3", []byte("value3"), 5*time.Second)
+
+	// All 3 should exist
+	_, found := cache.Get("key1")
+	suite.True(found)
+	_, found = cache.Get("key2")
+	suite.True(found)
+	_, found = cache.Get("key3")
+	suite.True(found)
+
+	// Add 4th entry - should evict oldest (key1)
+	cache.Set("key4", []byte("value4"), 5*time.Second)
+
+	suite.Equal(int64(3), cache.CountQueries())
+
+	// key1 should be evicted (it was least recently used)
+	_, found = cache.Get("key1")
+	suite.False(found)
+
+	// Others should still exist
+	_, found = cache.Get("key2")
+	suite.True(found)
+	_, found = cache.Get("key3")
+	suite.True(found)
+	_, found = cache.Get("key4")
+	suite.True(found)
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestLRUOrder() {
+	cache := NewLRUMemoryCache(1024*1024, 3) // Max 3 entries
+
+	cache.Set("key1", []byte("value1"), 5*time.Second)
+	cache.Set("key2", []byte("value2"), 5*time.Second)
+	cache.Set("key3", []byte("value3"), 5*time.Second)
+
+	// Access key1 to make it recently used
+	cache.Get("key1")
+
+	// Add key4 - should evict key2 (now least recently used)
+	cache.Set("key4", []byte("value4"), 5*time.Second)
+
+	// key2 should be evicted
+	_, found := cache.Get("key2")
+	suite.False(found)
+
+	// key1 should still exist (was accessed recently)
+	_, found = cache.Get("key1")
+	suite.True(found)
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestEvictionByMemory() {
+	// Small memory limit - 500 bytes
+	cache := NewLRUMemoryCache(500, 100)
+
+	// Each entry has ~64 bytes overhead + key + value
+	cache.Set("key1", []byte("value1"), 5*time.Second)
+	cache.Set("key2", []byte("value2"), 5*time.Second)
+	cache.Set("key3", []byte("value3"), 5*time.Second)
+
+	// Add large entry that should trigger eviction
+	largeValue := make([]byte, 200)
+	cache.Set("large", largeValue, 5*time.Second)
+
+	// Memory should be under limit
+	suite.LessOrEqual(cache.GetMemoryUsage(), int64(500))
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestCompression() {
+	cache := NewLRUMemoryCache(1024*1024, 100)
+
+	// Create a compressible value (> 1KB to trigger compression)
+	largeValue := make([]byte, 2048)
+	for i := range largeValue {
+		largeValue[i] = 'A' // Highly compressible
+	}
+
+	cache.Set("compressed", largeValue, 5*time.Second)
+
+	// Should be able to retrieve it correctly
+	val, found := cache.Get("compressed")
+	suite.True(found)
+	suite.Equal(largeValue, val)
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestGetStats() {
+	cache := NewLRUMemoryCache(1024*1024, 100)
+
+	cache.Set("key1", []byte("value1"), 5*time.Second)
+	cache.Set("key2", []byte("value2"), 5*time.Second)
+
+	stats := cache.GetStats()
+	suite.Equal(int64(2), stats["entries"])
+	suite.Equal(int64(1024*1024), stats["max_memory"])
+	suite.Equal(int64(100), stats["max_entries"])
+	suite.NotNil(stats["memory_bytes"])
+	suite.NotNil(stats["fill_percent"])
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestConcurrentAccess() {
+	cache := NewLRUMemoryCache(10*1024*1024, 1000)
+	const numGoroutines = 50
+	const numOperations = 500
+
+	var wg sync.WaitGroup
+	wg.Add(numGoroutines * 3) // readers, writers, deleters
+
+	// Writers
+	for i := 0; i < numGoroutines; i++ {
+		go func(id int) {
+			defer wg.Done()
+			for j := 0; j < numOperations; j++ {
+				key := fmt.Sprintf("key-%d-%d", id, j)
+				value := []byte(fmt.Sprintf("value-%d-%d", id, j))
+				cache.Set(key, value, 5*time.Second)
+			}
+		}(i)
+	}
+
+	// Readers
+	for i := 0; i < numGoroutines; i++ {
+		go func(id int) {
+			defer wg.Done()
+			for j := 0; j < numOperations; j++ {
+				key := fmt.Sprintf("key-%d-%d", id, j)
+				cache.Get(key)
+			}
+		}(i)
+	}
+
+	// Deleters
+	for i := 0; i < numGoroutines; i++ {
+		go func(id int) {
+			defer wg.Done()
+			for j := 0; j < numOperations; j++ {
+				key := fmt.Sprintf("key-%d-%d", id, j%100)
+				cache.Delete(key)
+			}
+		}(i)
+	}
+
+	wg.Wait()
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestCleanExpiredEntries() {
+	cache := NewLRUMemoryCache(1024*1024, 100)
+
+	cache.Set("expire1", []byte("value1"), 50*time.Millisecond)
+	cache.Set("expire2", []byte("value2"), 50*time.Millisecond)
+	cache.Set("keep", []byte("value3"), 5*time.Second)
+
+	suite.Equal(int64(3), cache.CountQueries())
+
+	// Wait for some to expire
+	time.Sleep(100 * time.Millisecond)
+
+	// Clean expired entries
+	cache.CleanExpiredEntries()
+
+	// Only "keep" should remain
+	suite.Equal(int64(1), cache.CountQueries())
+
+	_, found := cache.Get("keep")
+	suite.True(found)
+}
+
+func (suite *LRUMemoryCacheTestSuite) TestCountQueries() {
+	cache := NewLRUMemoryCache(1024*1024, 100)
+
+	suite.Equal(int64(0), cache.CountQueries())
+
+	cache.Set("key1", []byte("value1"), 5*time.Second)
+	suite.Equal(int64(1), cache.CountQueries())
+
+	cache.Set("key2", []byte("value2"), 5*time.Second)
+	suite.Equal(int64(2), cache.CountQueries())
+
+	cache.Delete("key1")
+	suite.Equal(int64(1), cache.CountQueries())
+
+	cache.Clear()
+	suite.Equal(int64(0), cache.CountQueries())
+}
+
+// Benchmarks
+
+func BenchmarkLRUMemoryCacheSet(b *testing.B) {
+	cache := NewLRUMemoryCache(100*1024*1024, 100000)
+	value := []byte("benchmark-value")
+
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		key := fmt.Sprintf("key-%d", i)
+		cache.Set(key, value, 5*time.Second)
+	}
+}
+
+func BenchmarkLRUMemoryCacheGet(b *testing.B) {
+	cache := NewLRUMemoryCache(100*1024*1024, 100000)
+	value := []byte("benchmark-value")
+
+	// Pre-populate
+	for i := 0; i < 10000; i++ {
+		key := fmt.Sprintf("key-%d", i)
+		cache.Set(key, value, 5*time.Minute)
+	}
+
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		key := fmt.Sprintf("key-%d", i%10000)
+		cache.Get(key)
+	}
+}
+
+func BenchmarkLRUMemoryCacheConcurrent(b *testing.B) {
+	cache := NewLRUMemoryCache(100*1024*1024, 100000)
+	value := []byte("benchmark-value")
+
+	b.RunParallel(func(pb *testing.PB) {
+		i := 0
+		for pb.Next() {
+			key := fmt.Sprintf("key-%d", i)
+			if i%2 == 0 {
+				cache.Set(key, value, 5*time.Second)
+			} else {
+				cache.Get(key)
+			}
+			i++
+		}
+	})
+}

errors.go

@@ -106,117 +106,8 @@ func (e *ProxyError) WithMetadata(key string, value interface{}) *ProxyError {
 	return e
 }
 
-// Common error constructors
-
-// NewConnectionError creates a connection-related error
-func NewConnectionError(err error) *ProxyError {
-	code := ErrCodeConnectionRefused
-	if err != nil {
-		errStr := err.Error()
-		if contains(errStr, "reset") {
-			code = ErrCodeConnectionReset
-		}
-	}
-
-	return NewProxyError(code, "Failed to connect to backend", 502, true).
-		WithCause(err)
-}
-
-// NewTimeoutError creates a timeout error
-func NewTimeoutError(err error) *ProxyError {
-	return NewProxyError(ErrCodeTimeout, "Request timed out", 504, false).
-		WithCause(err)
-}
-
-// NewCircuitOpenError creates a circuit breaker open error
-func NewCircuitOpenError() *ProxyError {
-	return NewProxyError(ErrCodeCircuitOpen, "Service temporarily unavailable due to circuit breaker", 503, false).
-		WithDetails("The backend service is currently experiencing issues. Please try again later.")
-}
-
-// NewRateLimitError creates a rate limit error
-func NewRateLimitError(userID, role string) *ProxyError {
-	return NewProxyError(ErrCodeRateLimited, "Rate limit exceeded", 429, false).
-		WithDetails("You have exceeded the rate limit for your role").
-		WithMetadata("user_id", userID).
-		WithMetadata("role", role)
-}
-
-// NewBackendError creates a backend error from status code
-func NewBackendError(statusCode int, body string) *ProxyError {
-	code := ErrCodeBackendError
-	message := "Backend returned an error"
-	retryable := false
-
-	switch {
-	case statusCode == 429:
-		code = ErrCodeRateLimited
-		message = "Backend rate limit exceeded"
-		retryable = true
-	case statusCode == 503:
-		code = ErrCodeServiceUnavailable
-		message = "Backend service unavailable"
-		retryable = true
-	case statusCode == 502 || statusCode == 504:
-		code = ErrCodeBadGateway
-		message = "Bad gateway"
-		retryable = true
-	case statusCode >= 500:
-		code = ErrCodeBackendError
-		message = "Backend server error"
-		retryable = true
-	case statusCode == 404:
-		code = ErrCodeNotFound
-		message = "Resource not found"
-	case statusCode == 403:
-		code = ErrCodeForbidden
-		message = "Access forbidden"
-	case statusCode == 401:
-		code = ErrCodeUnauthorized
-		message = "Unauthorized"
-	case statusCode >= 400:
-		code = ErrCodeInvalidRequest
-		message = "Invalid request"
-	}
-
-	return NewProxyError(code, message, statusCode, retryable).
-		WithMetadata("backend_status", statusCode).
-		WithMetadata("backend_body", truncateString(body, 500))
-}
-
-// NewInvalidResponseError creates an invalid response error
-func NewInvalidResponseError(details string) *ProxyError {
-	return NewProxyError(ErrCodeInvalidResponse, "Backend returned invalid response", 502, false).
-		WithDetails(details)
-}
-
-// NewInternalError creates an internal error
-func NewInternalError(err error) *ProxyError {
-	return NewProxyError(ErrCodeInternalError, "Internal proxy error", 500, false).
-		WithCause(err)
-}
-
-// NewContextCanceledError creates a context canceled error
-func NewContextCanceledError() *ProxyError {
-	return NewProxyError(ErrCodeContextCanceled, "Request canceled", 499, false).
-		WithDetails("The request was canceled by the client")
-}
-
 // Helper functions
 
-func contains(s, substr string) bool {
-	return len(s) > 0 && len(substr) > 0 && len(s) >= len(substr) && (s == substr || len(s) > len(substr) && (s[:len(substr)] == substr || s[len(s)-len(substr):] == substr || containsMiddle(s, substr)))
-}
-
-func containsMiddle(s, substr string) bool {
-	for i := 0; i <= len(s)-len(substr); i++ {
-		if s[i:i+len(substr)] == substr {
-			return true
-		}
-	}
-	return false
-}
-
 func truncateString(s string, maxLen int) string {
 	if len(s) <= maxLen {
 		return s

go.mod

@@ -20,7 +20,7 @@ require (
 	github.com/lukaszraczylo/ask v0.0.0-20240916204100-6e9ef53a62d9
 	github.com/lukaszraczylo/go-ratecounter v0.1.12
 	github.com/lukaszraczylo/go-simple-graphql v1.2.89
-	github.com/redis/go-redis/v9 v9.17.1
+	github.com/redis/go-redis/v9 v9.17.2
 	github.com/sony/gobreaker v1.0.0
 	github.com/stretchr/testify v1.11.1
 	github.com/valyala/fasthttp v1.68.0
@@ -47,7 +47,7 @@ require (
 	github.com/jackc/pgpassfile v1.0.0 // indirect
 	github.com/jackc/pgservicefile v0.0.0-20240606120523-5a60cdf6a761 // indirect
 	github.com/jackc/puddle/v2 v2.2.2 // indirect
-	github.com/klauspost/compress v1.18.1 // indirect
+	github.com/klauspost/compress v1.18.2 // indirect
 	github.com/mattn/go-colorable v0.1.14 // indirect
 	github.com/mattn/go-isatty v0.0.20 // indirect
 	github.com/mattn/go-runewidth v0.0.19 // indirect
@@ -67,8 +67,8 @@ require (
 	golang.org/x/sys v0.38.0 // indirect
 	golang.org/x/term v0.37.0 // indirect
 	golang.org/x/text v0.31.0 // indirect
-	google.golang.org/genproto/googleapis/api v0.0.0-20251124214823-79d6a2a48846 // indirect
-	google.golang.org/genproto/googleapis/rpc v0.0.0-20251124214823-79d6a2a48846 // indirect
+	google.golang.org/genproto/googleapis/api v0.0.0-20251202230838-ff82c1b0f217 // indirect
+	google.golang.org/genproto/googleapis/rpc v0.0.0-20251202230838-ff82c1b0f217 // indirect
 	google.golang.org/protobuf v1.36.10 // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect
 )

go.sum

@@ -64,8 +64,8 @@ github.com/jackc/pgx/v5 v5.7.6 h1:rWQc5FwZSPX58r1OQmkuaNicxdmExaEz5A2DO2hUuTk=
 github.com/jackc/pgx/v5 v5.7.6/go.mod h1:aruU7o91Tc2q2cFp5h4uP3f6ztExVpyVv88Xl/8Vl8M=
 github.com/jackc/puddle/v2 v2.2.2 h1:PR8nw+E/1w0GLuRFSmiioY6UooMp6KJv0/61nB7icHo=
 github.com/jackc/puddle/v2 v2.2.2/go.mod h1:vriiEXHvEE654aYKXXjOvZM39qJ0q+azkZFrfEOc3H4=
-github.com/klauspost/compress v1.18.1 h1:bcSGx7UbpBqMChDtsF28Lw6v/G94LPrrbMbdC3JH2co=
-github.com/klauspost/compress v1.18.1/go.mod h1:ZQFFVG+MdnR0P+l6wpXgIL4NTtwiKIdBnrBd8Nrxr+0=
+github.com/klauspost/compress v1.18.2 h1:iiPHWW0YrcFgpBYhsA6D1+fqHssJscY/Tm/y2Uqnapk=
+github.com/klauspost/compress v1.18.2/go.mod h1:R0h/fSBs8DE4ENlcrlib3PsXS61voFxhIs2DeRhCvJ4=
 github.com/kr/pretty v0.3.1 h1:flRD4NNwYAUpkphVc1HcthR4KEIFJ65n8Mw5qdRn3LE=
 github.com/kr/pretty v0.3.1/go.mod h1:hoEshYVHaxMs3cyo3Yncou5ZscifuDolrwPKZanG3xk=
 github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=
@@ -84,8 +84,8 @@ github.com/mattn/go-runewidth v0.0.19 h1:v++JhqYnZuu5jSKrk9RbgF5v4CGUjqRfBm05byF
 github.com/mattn/go-runewidth v0.0.19/go.mod h1:XBkDxAl56ILZc9knddidhrOlY5R/pDhgLpndooCuJAs=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
-github.com/redis/go-redis/v9 v9.17.1 h1:7tl732FjYPRT9H9aNfyTwKg9iTETjWjGKEJ2t/5iWTs=
-github.com/redis/go-redis/v9 v9.17.1/go.mod h1:u410H11HMLoB+TP67dz8rL9s6QW2j76l0//kSOd3370=
+github.com/redis/go-redis/v9 v9.17.2 h1:P2EGsA4qVIM3Pp+aPocCJ7DguDHhqrXNhVcEp4ViluI=
+github.com/redis/go-redis/v9 v9.17.2/go.mod h1:u410H11HMLoB+TP67dz8rL9s6QW2j76l0//kSOd3370=
 github.com/rogpeppe/go-internal v1.14.1 h1:UQB4HGPB6osV0SQTLymcB4TgvyWu6ZyliaW0tI/otEQ=
 github.com/rogpeppe/go-internal v1.14.1/go.mod h1:MaRKkUm5W0goXpeCfT7UZI6fk/L7L7so1lCWt35ZSgc=
 github.com/savsgio/gotils v0.0.0-20250924091648-bce9a52d7761 h1:McifyVxygw1d67y6vxUqls2D46J8W9nrki9c8c0eVvE=
@@ -144,10 +144,10 @@ golang.org/x/text v0.31.0 h1:aC8ghyu4JhP8VojJ2lEHBnochRno1sgL6nEi9WGFGMM=
 golang.org/x/text v0.31.0/go.mod h1:tKRAlv61yKIjGGHX/4tP1LTbc13YSec1pxVEWXzfoeM=
 gonum.org/v1/gonum v0.16.0 h1:5+ul4Swaf3ESvrOnidPp4GZbzf0mxVQpDCYUQE7OJfk=
 gonum.org/v1/gonum v0.16.0/go.mod h1:fef3am4MQ93R2HHpKnLk4/Tbh/s0+wqD5nfa6Pnwy4E=
-google.golang.org/genproto/googleapis/api v0.0.0-20251124214823-79d6a2a48846 h1:ZdyUkS9po3H7G0tuh955QVyyotWvOD4W0aEapeGeUYk=
-google.golang.org/genproto/googleapis/api v0.0.0-20251124214823-79d6a2a48846/go.mod h1:Fk4kyraUvqD7i5H6S43sj2W98fbZa75lpZz/eUyhfO0=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20251124214823-79d6a2a48846 h1:Wgl1rcDNThT+Zn47YyCXOXyX/COgMTIdhJ717F0l4xk=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20251124214823-79d6a2a48846/go.mod h1:7i2o+ce6H/6BluujYR+kqX3GKH+dChPTQU19wjRPiGk=
+google.golang.org/genproto/googleapis/api v0.0.0-20251202230838-ff82c1b0f217 h1:fCvbg86sFXwdrl5LgVcTEvNC+2txB5mgROGmRL5mrls=
+google.golang.org/genproto/googleapis/api v0.0.0-20251202230838-ff82c1b0f217/go.mod h1:+rXWjjaukWZun3mLfjmVnQi18E1AsFbDN9QdJ5YXLto=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20251202230838-ff82c1b0f217 h1:gRkg/vSppuSQoDjxyiGfN4Upv/h/DQmIR10ZU8dh4Ww=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20251202230838-ff82c1b0f217/go.mod h1:7i2o+ce6H/6BluujYR+kqX3GKH+dChPTQU19wjRPiGk=
 google.golang.org/grpc v1.77.0 h1:wVVY6/8cGA6vvffn+wWK5ToddbgdU3d8MNENr4evgXM=
 google.golang.org/grpc v1.77.0/go.mod h1:z0BY1iVj0q8E1uSQCjL9cppRj+gnZjzDnzV0dHhrNig=
 google.golang.org/protobuf v1.36.10 h1:AYd7cD/uASjIL6Q9LiTjz8JLcrh/88q5UObnmY3aOOE=

integration_test.go

@@ -1,6 +1,7 @@
 package main
 
 import (
+	"context"
 	"fmt"
 	"net/http"
 	"net/http/httptest"
@@ -783,3 +784,251 @@ func (suite *Tests) TestRequestCoalescingIntegration() {
 		}
 	})
 }
+
+// TestRetryBudgetIntegration tests that retry budget correctly limits retry attempts
+func (suite *Tests) TestRetryBudgetIntegration() {
+	// Initialize a retry budget with limited tokens for testing
+	budgetCtx := context.Background()
+	testBudget := NewRetryBudgetWithContext(budgetCtx, RetryBudgetConfig{
+		MaxTokens:       3, // Only allow 3 retries total
+		TokensPerSecond: 0, // Don't refill during test
+		Enabled:         true,
+	}, cfg.Logger)
+
+	// Replace global retry budget
+	originalBudget := retryBudget
+	retryBudget = testBudget
+	defer func() {
+		testBudget.Shutdown()
+		retryBudget = originalBudget
+	}()
+
+	suite.Run("retry_budget_limits_retries", func() {
+		testBudget.Reset()
+
+		// Verify retry budget is set and works correctly
+		rb := GetRetryBudget()
+		suite.NotNil(rb, "Retry budget should be set")
+		suite.True(rb.enabled, "Retry budget should be enabled")
+		suite.T().Logf("Retry budget: enabled=%v, tokens=%d", rb.enabled, rb.currentTokens.Load())
+
+		// Test that AllowRetry consumes tokens correctly
+		initialTokens := rb.currentTokens.Load()
+		suite.Equal(int64(3), initialTokens, "Should start with 3 tokens")
+
+		// First 3 retries should be allowed
+		suite.True(rb.AllowRetry(), "First retry should be allowed")
+		suite.True(rb.AllowRetry(), "Second retry should be allowed")
+		suite.True(rb.AllowRetry(), "Third retry should be allowed")
+
+		// Fourth retry should be denied (tokens exhausted)
+		suite.False(rb.AllowRetry(), "Fourth retry should be denied - budget exhausted")
+
+		// Verify stats
+		stats := rb.GetStats()
+		suite.Equal(int64(4), stats["total_attempts"].(int64), "Should have 4 total attempts")
+		suite.Equal(int64(3), stats["allowed_retries"].(int64), "Should have 3 allowed retries")
+		suite.Equal(int64(1), stats["denied_retries"].(int64), "Should have 1 denied retry")
+
+		suite.T().Logf("Retry budget stats: total=%d, allowed=%d, denied=%d",
+			stats["total_attempts"], stats["allowed_retries"], stats["denied_retries"])
+	})
+
+	suite.Run("retry_budget_exhaustion", func() {
+		// Create a new budget with only 1 token
+		testBudget.Shutdown()
+		budgetCtx2 := context.Background()
+		testBudget2 := NewRetryBudgetWithContext(budgetCtx2, RetryBudgetConfig{
+			MaxTokens:       1, // Only allow 1 retry
+			TokensPerSecond: 0, // Don't refill
+			Enabled:         true,
+		}, cfg.Logger)
+		retryBudget = testBudget2
+		defer func() {
+			testBudget2.Shutdown()
+		}()
+
+		// Test budget exhaustion with 1 token
+		rb := GetRetryBudget()
+		suite.NotNil(rb, "Retry budget should be set")
+		suite.Equal(int64(1), rb.currentTokens.Load(), "Should start with 1 token")
+
+		// First retry should be allowed
+		suite.True(rb.AllowRetry(), "First retry should be allowed")
+
+		// Second retry should be denied (only 1 token available)
+		suite.False(rb.AllowRetry(), "Second retry should be denied - budget exhausted")
+
+		// Verify stats
+		stats := rb.GetStats()
+		suite.Equal(int64(2), stats["total_attempts"].(int64), "Should have 2 total attempts")
+		suite.Equal(int64(1), stats["allowed_retries"].(int64), "Should have 1 allowed retry")
+		suite.Equal(int64(1), stats["denied_retries"].(int64), "Should have 1 denied retry")
+
+		suite.T().Logf("Retry budget stats: total=%d, allowed=%d, denied=%d",
+			stats["total_attempts"], stats["allowed_retries"], stats["denied_retries"])
+	})
+}
+
+// TestConnectionPoolStatsIntegration tests that connection pool stats are tracked
+func (suite *Tests) TestConnectionPoolStatsIntegration() {
+	// Save original config
+	originalClient := cfg.Client.FastProxyClient
+	originalHostGraphQL := cfg.Server.HostGraphQL
+	originalCoalescing := cfg.RequestCoalescing.Enable
+
+	// Restore after test
+	defer func() {
+		cfg.Client.FastProxyClient = originalClient
+		cfg.Server.HostGraphQL = originalHostGraphQL
+		cfg.RequestCoalescing.Enable = originalCoalescing
+	}()
+
+	// Disable request coalescing for accurate tracking
+	cfg.RequestCoalescing.Enable = false
+
+	suite.Run("connection_success_tracked", func() {
+		// Create test server that succeeds
+		server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+			w.Header().Set("Content-Type", "application/json")
+			w.WriteHeader(http.StatusOK)
+			_, _ = w.Write([]byte(`{"data":{"test":"success"}}`))
+		}))
+		defer server.Close()
+
+		cfg.Server.HostGraphQL = server.URL
+		cfg.Client.ClientTimeout = 5
+		cfg.Client.FastProxyClient = createFasthttpClient(cfg)
+
+		// Initialize connection pool
+		InitializeConnectionPool(cfg.Client.FastProxyClient)
+		defer ShutdownConnectionPool()
+
+		poolMgr := GetConnectionPoolManager()
+		suite.NotNil(poolMgr, "Connection pool manager should be initialized")
+
+		// Get stats before
+		statsBefore := poolMgr.GetConnectionStats()
+		successBefore := statsBefore["total_connections"].(int64)
+
+		// Make a successful request
+		reqCtx := &fasthttp.RequestCtx{}
+		reqCtx.Request.SetRequestURI("/graphql")
+		reqCtx.Request.Header.SetMethod("POST")
+		reqCtx.Request.Header.Set("Content-Type", "application/json")
+		reqCtx.Request.SetBody([]byte(`{"query": "query { test }"}`))
+
+		ctx := suite.app.AcquireCtx(reqCtx)
+		err := proxyTheRequest(ctx, cfg.Server.HostGraphQL)
+		suite.app.ReleaseCtx(ctx)
+
+		suite.Nil(err, "Request should succeed")
+
+		// Get stats after
+		statsAfter := poolMgr.GetConnectionStats()
+		successAfter := statsAfter["total_connections"].(int64)
+
+		suite.Greater(successAfter, successBefore,
+			"Total connections should increase after successful request")
+	})
+
+	suite.Run("connection_failure_tracked_on_5xx", func() {
+		// Create test server that returns 503
+		// Note: 503 triggers retry which records failures
+		server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+			w.WriteHeader(http.StatusServiceUnavailable)
+			_, _ = w.Write([]byte(`{"errors":[{"message":"Service unavailable"}]}`))
+		}))
+		defer server.Close()
+
+		cfg.Server.HostGraphQL = server.URL
+		cfg.Client.ClientTimeout = 2
+		cfg.Client.FastProxyClient = createFasthttpClient(cfg)
+
+		// Initialize connection pool
+		InitializeConnectionPool(cfg.Client.FastProxyClient)
+		defer ShutdownConnectionPool()
+
+		poolMgr := GetConnectionPoolManager()
+		suite.NotNil(poolMgr, "Connection pool manager should be initialized")
+
+		// Get stats before
+		statsBefore := poolMgr.GetConnectionStats()
+		failuresBefore := statsBefore["connection_failures"].(int64)
+
+		// Make a failing request (503 is retryable, so it will retry and track failures)
+		reqCtx := &fasthttp.RequestCtx{}
+		reqCtx.Request.SetRequestURI("/graphql")
+		reqCtx.Request.Header.SetMethod("POST")
+		reqCtx.Request.Header.Set("Content-Type", "application/json")
+		reqCtx.Request.SetBody([]byte(`{"query": "query { fail }"}`))
+
+		ctx := suite.app.AcquireCtx(reqCtx)
+		_ = proxyTheRequest(ctx, cfg.Server.HostGraphQL)
+		suite.app.ReleaseCtx(ctx)
+
+		// Get stats after - should have failures from retry attempts
+		statsAfter := poolMgr.GetConnectionStats()
+		failuresAfter := statsAfter["connection_failures"].(int64)
+
+		suite.Greater(failuresAfter, failuresBefore,
+			"Connection failures should increase after 5xx responses that trigger retries")
+
+		suite.T().Logf("Connection failures: before=%d, after=%d",
+			failuresBefore, failuresAfter)
+	})
+
+	suite.Run("stats_reflect_request_outcomes", func() {
+		// This test verifies that connection stats properly reflect the
+		// combination of successes and failures over multiple requests
+
+		// Start with a fresh server
+		var requestCount atomic.Int32
+		server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+			count := requestCount.Add(1)
+			// First 2 requests succeed, rest fail
+			if count <= 2 {
+				w.Header().Set("Content-Type", "application/json")
+				w.WriteHeader(http.StatusOK)
+				_, _ = w.Write([]byte(`{"data":{"test":"success"}}`))
+			} else {
+				w.WriteHeader(http.StatusInternalServerError)
+				_, _ = w.Write([]byte(`{"errors":[{"message":"Error"}]}`))
+			}
+		}))
+		defer server.Close()
+
+		cfg.Server.HostGraphQL = server.URL
+		cfg.Client.ClientTimeout = 2
+		cfg.Client.FastProxyClient = createFasthttpClient(cfg)
+
+		// Initialize connection pool
+		InitializeConnectionPool(cfg.Client.FastProxyClient)
+		defer ShutdownConnectionPool()
+
+		poolMgr := GetConnectionPoolManager()
+		suite.NotNil(poolMgr, "Connection pool manager should be initialized")
+
+		// Make 2 successful requests
+		for i := 0; i < 2; i++ {
+			reqCtx := &fasthttp.RequestCtx{}
+			reqCtx.Request.SetRequestURI("/graphql")
+			reqCtx.Request.Header.SetMethod("POST")
+			reqCtx.Request.Header.Set("Content-Type", "application/json")
+			reqCtx.Request.SetBody([]byte(`{"query": "query { test }"}`))
+
+			ctx := suite.app.AcquireCtx(reqCtx)
+			_ = proxyTheRequest(ctx, cfg.Server.HostGraphQL)
+			suite.app.ReleaseCtx(ctx)
+		}
+
+		// Get stats after successes
+		statsAfterSuccess := poolMgr.GetConnectionStats()
+		totalConnections := statsAfterSuccess["total_connections"].(int64)
+
+		suite.GreaterOrEqual(totalConnections, int64(2),
+			"Should have at least 2 successful connections tracked")
+
+		suite.T().Logf("Total connections after 2 successful requests: %d", totalConnections)
+	})
+}

main.go

@@ -131,6 +131,7 @@ func parseConfig() {
 	c.Cache.CacheTTL = getDetailsFromEnv("CACHE_TTL", 60)
 	c.Cache.CacheMaxMemorySize = getDetailsFromEnv("CACHE_MAX_MEMORY_SIZE", 100) // Default 100MB
 	c.Cache.CacheMaxEntries = getDetailsFromEnv("CACHE_MAX_ENTRIES", 10000)      // Default 10000 entries
+	c.Cache.CacheUseLRU = getDetailsFromEnv("CACHE_USE_LRU", false)              // Use LRU eviction algorithm
 	// GraphQL query parsing cache - auto-calculate based on CPU cores if not set
 	c.Cache.GraphQLQueryCacheSize = getDetailsFromEnv("GRAPHQL_QUERY_CACHE_SIZE", runtime.GOMAXPROCS(0)*250)
 
@@ -390,9 +391,16 @@ func parseConfig() {
 			// Memory cache configurations
 			cacheConfig.Memory.MaxMemorySize = int64(cfg.Cache.CacheMaxMemorySize) * 1024 * 1024 // Convert MB to bytes
 			cacheConfig.Memory.MaxEntries = int64(cfg.Cache.CacheMaxEntries)
+			cacheConfig.Memory.UseLRU = cfg.Cache.CacheUseLRU
+
+			cacheType := "standard"
+			if cfg.Cache.CacheUseLRU {
+				cacheType = "LRU"
+			}
 			cfg.Logger.Info(&libpack_logging.LogMessage{
 				Message: "Configuring memory cache with limits",
 				Pairs: map[string]interface{}{
+					"type":          cacheType,
 					"max_memory_mb": cfg.Cache.CacheMaxMemorySize,
 					"max_entries":   cfg.Cache.CacheMaxEntries,
 				},

metrics_aggregator.go

@@ -519,7 +519,10 @@ func (ma *MetricsAggregator) aggregateStats(instances []InstanceMetrics) map[str
 		totalRetryAllowed  int64
 		totalRetryDenied   int64
 		totalRetryAttempts int64
+		totalCurrentTokens int64
+		totalMaxTokens     int64
 		retryBudgetEnabled = false
+		retryTokensPerSec  float64 // Use max tokens_per_sec from any instance
 
 		// Circuit breaker stats
 		cbOpenCount           int
@@ -645,6 +648,17 @@ func (ma *MetricsAggregator) aggregateStats(instances []InstanceMetrics) map[str
 			if attempts, ok := instance.RetryBudget["total_attempts"].(float64); ok {
 				totalRetryAttempts += int64(attempts)
 			}
+			if currentTokens, ok := instance.RetryBudget["current_tokens"].(float64); ok {
+				totalCurrentTokens += int64(currentTokens)
+			}
+			if maxTokens, ok := instance.RetryBudget["max_tokens"].(float64); ok {
+				totalMaxTokens += int64(maxTokens)
+			}
+			if tokensPerSec, ok := instance.RetryBudget["tokens_per_sec"].(float64); ok {
+				if tokensPerSec > retryTokensPerSec {
+					retryTokensPerSec = tokensPerSec
+				}
+			}
 		}
 
 		// Aggregate circuit breaker stats
@@ -748,6 +762,9 @@ func (ma *MetricsAggregator) aggregateStats(instances []InstanceMetrics) map[str
 			"denied_retries":  totalRetryDenied,
 			"total_attempts":  totalRetryAttempts,
 			"denial_rate_pct": retryDenialRate,
+			"current_tokens":  totalCurrentTokens,
+			"max_tokens":      totalMaxTokens,
+			"tokens_per_sec":  retryTokensPerSec,
 		},
 		"circuit_breaker": map[string]interface{}{
 			"enabled":            circuitBreakerEnabled,

monitoring/defaults.go

@@ -9,7 +9,3 @@ func (ms *MetricsSetup) RegisterDefaultMetrics() {
 	ms.RegisterMetricsCounter(MetricsCacheMiss, nil)
 	ms.RegisterMetricsCounter(MetricsQueriesCached, nil)
 }
-
-func (ms *MetricsSetup) RegisterGoMetrics() {
-	// TODO: metrics.WriteProcessMetrics(ms.metrics_set)
-}

proxy.go

@@ -454,19 +454,36 @@ func executeProxyAttempt(c *fiber.Ctx, proxyURL string) error {
 		return retry.Unrecoverable(fmt.Errorf("fiber context became nil during retry"))
 	}
 
+	// Get connection pool manager for stats tracking
+	poolMgr := GetConnectionPoolManager()
+
 	// Execute the proxy request
-	if err := doProxyRequestWithTimeout(c, proxyURL, cfg.Client.FastProxyClient); err != nil {
+	proxyErr := doProxyRequestWithTimeout(c, proxyURL, cfg.Client.FastProxyClient)
+	if proxyErr != nil {
 		// Check if this is a connection error
-		if isConnectionError(err) {
+		if isConnectionError(proxyErr) {
 			notifyHealthManager(false)
-			return err // Connection errors are retryable
+			// Track connection failure
+			if poolMgr != nil {
+				poolMgr.RecordConnectionFailure()
+			}
+			return proxyErr // Connection errors are retryable
 		}
 
 		// Check if this is a timeout error - don't retry timeouts
-		if isTimeoutError(err) {
-			return retry.Unrecoverable(err)
+		if isTimeoutError(proxyErr) {
+			return retry.Unrecoverable(proxyErr)
 		}
-		return err
+
+		// Check if this is a retryable HTTP error (e.g., 503)
+		// These indicate the server responded but with an error status
+		if strings.Contains(proxyErr.Error(), "non-200 response") {
+			// Track as a failure for retryable HTTP errors
+			if poolMgr != nil {
+				poolMgr.RecordConnectionFailure()
+			}
+		}
+		return proxyErr
 	}
 
 	// Safety check before accessing response
@@ -481,10 +498,18 @@ func executeProxyAttempt(c *fiber.Ctx, proxyURL string) error {
 	if err == nil {
 		// Success case
 		notifyHealthManager(true)
+		// Track successful connection
+		if poolMgr != nil {
+			poolMgr.RecordConnectionSuccess()
+		}
 		return nil
 	}
 
 	if shouldRetry {
+		// Track connection failure for retryable errors (5xx, etc)
+		if poolMgr != nil {
+			poolMgr.RecordConnectionFailure()
+		}
 		return err // Retryable error
 	}
 
@@ -541,31 +566,51 @@ func performProxyRequestWithEnhancedRetries(c *fiber.Ctx, proxyURL string, backe
 		retry.LastErrorOnly(true),
 		retry.RetryIf(func(err error) bool {
 			// Don't retry if context is cancelled or context is nil
-			defer func() {
-				// Recover from any panic when accessing context
-				if r := recover(); r != nil {
-					// If we panic, don't retry
-					return
-				}
-			}()
-
 			if c == nil {
 				return false
 			}
 
-			// Try to safely access the context
-			ctx := c.Context()
-			if ctx == nil {
+			// Safely check if context is done/cancelled
+			// Note: fasthttp.RequestCtx.Done() can panic if not properly initialized
+			// If we panic, don't retry (maintains backward compatibility with test behavior)
+			shouldRetry := true
+			func() {
+				defer func() {
+					if r := recover(); r != nil {
+						// If we panic accessing context, don't retry
+						// This typically happens in test scenarios with mock contexts
+						shouldRetry = false
+					}
+				}()
+				ctx := c.Context()
+				if ctx == nil {
+					return
+				}
+				select {
+				case <-ctx.Done():
+					shouldRetry = false
+				default:
+				}
+			}()
+
+			if !shouldRetry {
 				return false
 			}
 
-			// Check if context is done/cancelled
-			select {
-			case <-ctx.Done():
-				return false
-			default:
-				return true
+			// Check retry budget before allowing retry
+			if rb := GetRetryBudget(); rb != nil {
+				if !rb.AllowRetry() {
+					cfg.Logger.Warning(&libpack_logger.LogMessage{
+						Message: "Retry denied by budget",
+						Pairs: map[string]interface{}{
+							"path":  c.Path(),
+							"error": err.Error(),
+						},
+					})
+					return false
+				}
 			}
+			return true
 		}),
 	)
 }

struct_config.go

@@ -44,6 +44,7 @@ type config struct {
 		CacheRedisEnable      bool
 		CacheMaxMemorySize    int
 		CacheMaxEntries       int
+		CacheUseLRU           bool // Use LRU eviction algorithm instead of random eviction
 		GraphQLQueryCacheSize int  // Max number of parsed GraphQL queries to cache
 		PerUserCacheDisabled  bool // Disable per-user cache isolation (SECURITY RISK - not recommended)
 	}